hashirama/uvg266
1
0
Fork 0
mirror of https://github.com/ultravideo/uvg266.git synced 2024-11-24 18:34:06 +00:00
Code Issues Actions Projects Releases Wiki Activity
2413 commits 3 branches 6 tags 9.2 MiB
Commit graph

2091 commits

Author SHA1 Message Date
Ari Koivula a7a33b08ec Remove --slice-addresses from usage message 
And give a warning if it's used.

Slices will have to be implemented at some point, but they aren't yet
so let's not advertize them.
 2016-09-10 21:06:00 +03:00
Eemeli Kallio f41e428e5f Removed kvz_skip_unnecessary_rdoq and reworked --rdoq-skip to skip 4x4 blocks when it is on. 2016-09-09 10:26:07 +03:00
Eemeli Kallio ed9c0b0416 RDOQ reworked in rdo.c. rdoq_signhide now skips coeffs that are after best_last_idx. 2016-09-09 10:16:51 +03:00
Ari Koivula 02cd17b427 Add faster AVX inter SAD for 32x32 and 64x64 
Add implementations for these functions that process the image line by
line instead of using the 16x16 function to process block by block.

The 32x32 is around 30% faster, and 64x64 is around 15% faster,
on Haswell.

PASS inter_sad: 28.744M x reg_sad(32x32):x86_asm_avx (1014 ticks, 1.014 sec)
PASS inter_sad: 7.882M x reg_sad(64x64):x86_asm_avx (1014 ticks, 1.014 sec)
to
PASS inter_sad: 37.828M x reg_sad(32x32):x86_asm_avx (1014 ticks, 1.014 sec)
PASS inter_sad: 9.081M x reg_sad(64x64):x86_asm_avx (1014 ticks, 1.014 sec)
 2016-09-01 21:36:39 +03:00
Ari Koivula d0512d25c6 Use fixed point in get_mvd_coding_cost 2016-08-30 21:37:12 +03:00
Ari Koivula ec7507a935 Further optimize get_ep_ex_golomb_bitcost 
Unrolled 16-bit log2 calculation.
 2016-08-30 21:37:01 +03:00
Ari Koivula a4ba794587 Optimize get_ep_ex_golomb_bitcost 
Arrange the decision tree such that there is only 3 branches on the
most common paths and the more likely branch is always fall-through.

A profile guided optimization pass would probably do something similar.
 2016-08-30 05:24:16 +03:00
Ari Koivula 82cfab58f8 Improve fast mvd coding cost estimation 
A lot of time is being taken up by this function on ultrafast, and it
doesn't do a very good job. This change aims to both simplify the
logic and make the estimate better.

The logic is simplified by using a look up for the step mvd bit cost
step function instead of mimicking the binarization process. The
estimation is made better by checking fractional cabac bit costs.

The new function returns the same results as
kvz_get_mvd_coding_cost_cabac, but is also faster than the old
function.
 2016-08-30 04:55:09 +03:00
Ari Koivula d31be8eb27 Make mvd_coding_cost functions take const cabac 2016-08-30 04:46:46 +03:00
Ari Koivula 64d631c174 Fix 8bit to 10bit input conversion regression 2016-08-25 22:09:40 +03:00
Ari Koivula 27789125d8 Fix input bit depth conversion 
The input was being shifted to the wrong direction.
 2016-08-25 22:05:25 +03:00
Ari Koivula 4ec039004b Add monochrome encoding 
Write bitstream without chroma when encoding with --input-format=P400.
This reduces bitstream size by 0-1 %, compared to coding monochrome in
420 format, and speeds up encoding slightly due to not processing
chroma.
 2016-08-25 20:15:26 +03:00
Ari Koivula c5b70cf812 Add chroma format support to yuv_t 2016-08-24 19:20:53 +03:00
Ari Koivula 032ed30ff4 Add chroma format support to kvz_picture 
Add picture_alloc_csp to libkvz api to allocated pictures with chroma
format different from 420.
 2016-08-24 19:20:53 +03:00
Ari Koivula 48ccc26839 Add --input-format and --input-bitdepth 
Adds reading of 10 bit input for 10-bit encoding.
 2016-08-24 19:20:53 +03:00
Ari Koivula cc08073615 Refactor some indexing weirdness in init_lcu_t 
I thought there might be a bug in this so I cleaned it up.
 2016-08-24 19:12:48 +03:00
Ari Koivula b6d674d66e Refactor integer vector inter prediction 
This code was pretty bad, so I cleaned it up a bit.
 2016-08-24 19:09:26 +03:00
Ari Lemmetti 28c4174d0e Fix incorrect shuffle parameters 
_MM_SHUFFLE uses reverse order
 2016-08-23 19:40:46 +03:00
Ari Lemmetti ce77bfa15b Replace KVZ_PERMUTE with _MM_SHUFFLE 
The same exact macro already exists
 2016-08-22 19:08:46 +03:00
Jovasa 68eef660bd Fixed search around mv_in in fullsearch not being saved. 2016-08-19 15:19:29 +03:00
Eemeli Kallio 99d8b9abeb Changed skip_rdoq name to kvz_skip_unnecessary_rdoq. Changed the order it uses when it goes through CGs and tuned its sum calculation. 2016-08-18 14:02:56 +03:00
Eemeli Kallio 1fb4755f31 Added rdoq-skip to quant-generic.c 2016-08-18 12:17:54 +03:00
Eemeli Kallio d20ac03ca2 Added --rdoq-skip option 2016-08-18 12:17:53 +03:00
Marko Viitanen 83cf801664 Fixed MV constraint condition in bipred 2016-08-18 08:53:17 +03:00
Marko Viitanen 5ae1c595f2 Fixed slice_temporal_mvp_enabled_flag and disabled TMVP with tiles 
- slice_temporal_mvp_enabled_flag should be signalled also with non-IDR I-slices
 2016-08-10 14:51:41 +03:00
Marko Viitanen 5326519182 TMVP cleanup and const qualifier fixes 2016-08-10 14:10:43 +03:00
Marko Viitanen f40907260d Added config parameter for TMVP and cmdline option --no-tmvp 
- Enabled by default
 - Cannot be used with GOP at the moment
 2016-08-10 14:09:29 +03:00
Marko Viitanen fd52dac1f7 Fixed TMVP scaling 2016-08-10 14:09:28 +03:00
Marko Viitanen c664bc8cf7 Added flag collocated_ref_idx to the slice header 2016-08-10 14:09:28 +03:00
Marko Viitanen c5f2611a38 Fixes for TMVP to work with the new CU array 2016-08-10 14:09:28 +03:00
Marko Viitanen d85af5755b TMVP working when only 1 ref frame 2016-08-10 14:09:28 +03:00
Marko Viitanen 39f0165efe Fix a bug in TMVP, the reference cu_array was being overwritten 2016-08-10 14:09:27 +03:00
Marko Viitanen adab8c327e Clean TMVP code 2016-08-10 14:09:20 +03:00
Marko Viitanen 5fa8226ac9 Temporal merge candidate selection 2016-08-10 14:09:20 +03:00
Marko Viitanen f83042f4a1 Temporal MV candidate selection 2016-08-10 14:09:19 +03:00
Marko Viitanen f8671581e3 Implemented function kvz_inter_get_temporal_merge_candidates() 2016-08-10 14:09:19 +03:00
Marko Viitanen 2956bdb379 Added flag slice_temporal_mvp_enabled_flag 2016-08-10 14:09:19 +03:00
Arttu Ylä-Outinen 2a946bd88e Rename encoder_state_t.global to frame 
"Frame" is more accurate than "global" since when OWF is used, encoder
states for each frame have their own struct.
 2016-08-10 13:22:36 +09:00
Arttu Ylä-Outinen 5fbb0a8c27 Fix includes 2016-08-10 13:05:40 +09:00
Arttu Ylä-Outinen aabf6ca3ee Extract encoding code from encoderstate.c 
Moves functions kvz_encode_coding_tree and kvz_encode_coeff_nxn from
encoderstate.c to encode_coding_tree.c.
 2016-08-09 22:16:50 +09:00
Arttu Ylä-Outinen 803f29be8f Remove reconstructed picture allocation in lossless. 
Changes encoder_set_source_picture to set the reconstructed picture to
a copy of the source picture instead of allocating a new picture when
lossless coding is used.
 2016-08-03 14:25:08 +09:00
Arttu Ylä-Outinen aaec473a19 Refactor encoder state initialization. 
- Moves allocation of the reconstructed picture after the source picture
  is set.
- Extracts main state initialization to a separate function from
  encoder_state_new_frame.
- Changes kvz_encoder_feed_frame to return the frame.
- Renames some functions to better match their purpose.
 2016-08-03 14:25:08 +09:00
Arttu Ylä-Outinen cd7024b3a5 Skip computing SSD when using lossless coding. 
The SSD is always zero since it is lossless.
 2016-08-03 14:25:08 +09:00
Arttu Ylä-Outinen fbbe5d1844 Use kvz_pixels_calc_ssd for SSD in search.c. 
Replaces loops for computing SSDs by calling kvz_pixels_calc_ssd in
search.c.
 2016-08-03 14:25:08 +09:00
Arttu Ylä-Outinen 22cc97ffb1 Fix missing field initializers. 2016-08-03 14:25:08 +09:00
Arttu Ylä-Outinen 06b82bf888 Disable filters, trskip and signhide in lossless. 
When lossless coding is used, deblock and SAO are skipped, transform
skip flag is not written and sign hiding is not used.
 2016-08-03 14:25:08 +09:00
Arttu Ylä-Outinen 97451ec401 Align assignments in encoder.c. 2016-08-03 14:25:08 +09:00
Arttu Ylä-Outinen 1dc94663c3 Bypass transform and quantization with --lossless. 
When --lossless is given, set cu_transquant_bypass_flag for every CU and
bypass transform and quantization by directly copying reference pixels
to reconstruction and the residual to coefficients.
 2016-08-03 14:25:08 +09:00
Arttu Ylä-Outinen 2113b0182d Enable PPS-level tq bypass flag with --lossless. 
Sets transquant_bypass_enable_flag to true in PPS when --lossless is
given.
 2016-08-03 14:25:08 +09:00
Arttu Ylä-Outinen a5897bbece Make cabac context initialization tables static. 2016-08-03 14:25:08 +09:00
Arttu Ylä-Outinen 23e7d9bb37 Add --lossless command line parameter. 2016-08-03 14:25:08 +09:00
Arttu Ylä-Outinen 5372ea432f Update README and manpage. 2016-08-03 14:25:08 +09:00
Ari Lemmetti 6bcba004ff Comment out to fix unused code error on clang. 2016-07-14 14:12:16 +03:00
Ari Lemmetti c0979ebdcb Implement AVX2 luma sampling 2016-07-14 12:53:02 +03:00
Ari Lemmetti 6244560426 Add avx2 strategy for kvz_filter_frac_blocks_luma. 2016-07-14 12:53:02 +03:00
Ari Lemmetti 9c4e9e049b Load only what is needed. Eliminate latency from hadds. 2016-07-14 12:53:01 +03:00
Ari Lemmetti 7f71cb423a Check 4 fractional pixel positions simultaneously 2016-07-14 12:52:24 +03:00
Ari Lemmetti ad445ab8a1 Transition to kvz_filter_frac_blocks_luma 2016-07-14 12:51:02 +03:00
Ari Lemmetti fccfbd2f28 Add strategy for kvz_filter_frac_blocks_luma 2016-07-14 12:51:02 +03:00
Ari Lemmetti e9c3074d32 Add buffers and definitions for upcoming filtering 
Samples are to be filtered in separate blocks instead of
making one big picture with interpolated pixels
 2016-07-14 12:51:02 +03:00
Ari Lemmetti 7afe7e963b Use fme_level to control the search accuracy. 2016-07-14 12:51:01 +03:00
Ari Lemmetti 5fa323bf25 Skip searching best hpel twice. Make hpel and qpel loops similar. 2016-07-14 12:51:01 +03:00
Ari Lemmetti bc98a9affa Change the search order to suit lighter fme search 2016-07-14 12:51:01 +03:00
Ari Lemmetti 2b0c8db349 Add quad satd for avx2 2016-07-14 12:50:24 +03:00
Ari Lemmetti 0ff69fd6f8 Add any size multi satd 2016-07-14 12:48:37 +03:00
Ari Lemmetti d17b9e7d6e Allow subme parameters 0-4 
Update usage, presets,defaults,lib version
 2016-07-12 19:49:38 +03:00
Arttu Ylä-Outinen 62ad57d0bf Fix kvz_image_list_add for zero-sized lists. 
When a list does not have space for the new element, its size is
doubled. If the size of the list is zero, it would not be resized. Fixed
to always resize the list so that the new element can be added.
 2016-06-22 13:35:16 +09:00
Arttu Ylä-Outinen 433e528af7 Drop unused variable in search_pu_inter. 
Removes unused variable max_px_below_lcu.
 2016-06-22 13:35:16 +09:00
Arttu Ylä-Outinen 7836ff6ec9 Drop unused functions. 
Removes functions kvz_coefficients_calc_abs, kvz_intra_rdo_cost_compare
and kvz_rdo_cost_intra which are no longer used.
 2016-06-22 13:35:15 +09:00
Arttu Ylä-Outinen e4b5840f56 Add parentheses around macro arguments in cabac.h. 2016-06-22 13:35:15 +09:00
Arttu Ylä-Outinen a387b74e51 Fix resolution auto-detection. 
Only try to guess the resolution from filename when neither width nor
height is given.
 2016-06-22 13:35:15 +09:00
Arttu Ylä-Outinen 097bf8f3c0 Add a typedef for mvd coding cost functions. 2016-06-20 13:56:10 +09:00
Arttu Ylä-Outinen d3c0e49286 Update comments. 2016-06-16 20:25:08 +09:00
Arttu Ylä-Outinen ae832cda8c Pack cbf flags in cu_info_t to two bytes. 
Reduces size of cu_info_t.
 2016-06-16 20:24:19 +09:00
Arttu Ylä-Outinen cad2d496b8 Enable 4x8 and 4x16 partition modes 
Enables search for 2NxN and Nx2N partition modes for 8x8 CUs and 2NxnU,
2NxnD, nLx2N and nRx2N partition modes for 16x16 CUs.

Changes the loop for copying reconstructed luma pixels in
kvz_inter_recon_lcu to use 4 byte chunks instead of 8 byte chunks since
it is now possible to have 4 pixel wide blocks.
 2016-06-16 20:23:16 +09:00
Arttu Ylä-Outinen 90df7350f0 Make deblocking work with 4 pixel wide blocks. 2016-06-16 20:21:50 +09:00
Arttu Ylä-Outinen bf26661782 Add support for 4x4 blocks to SATD_ANY_SIZE. 
Makes functions satd_any_size_generic and satd_any_size_8bit_avx2 work
on blocks whose width and/or height are not multiples of 8.
 2016-06-16 18:53:17 +09:00
Arttu Ylä-Outinen 2ae260e422 Change width of cells in lcu_t to 4 pixels. 
Intra mode info for NxN partition units is now stored in the
corresponding 4x4 cell in lcu_t.cu array.
 2016-06-16 18:53:17 +09:00
Arttu Ylä-Outinen 360f5bb8da Always use pixel coordinates for indexing lcu_t. 
Removes macro LCU_GET_CU and uses LCU_GET_CU_AT_PX in its place.
 2016-06-16 18:53:17 +09:00
Arttu Ylä-Outinen 46e8122d27 Add functions for indexing cu_array_t structures. 
Replaces macro CU_ARRAY_AT with functions kvz_cu_array_at and
kvz_cu_array_at_const.
 2016-06-16 18:52:19 +09:00
Arttu Ylä-Outinen c5afabdd3b Change width of cells in cu_array_t to 4 pixels. 2016-06-15 12:25:11 +09:00
Arttu Ylä-Outinen 57a3d9b4b9 Add a function for copying CU data from LCUs. 
Adds function kvz_cu_array_copy_from_lcu which CU info data from an
lcu_t structure to a cu_array_t structure.
 2016-06-15 12:25:11 +09:00
Arttu Ylä-Outinen 2c85a00a55 Change kvz_cu_array_alloc to use pixel dimensions. 
Changes function kvz_cu_array_alloc to take width and height parameters
in pixels instead of SCUs.
 2016-06-15 12:25:11 +09:00
Arttu Ylä-Outinen b276a347c0 Add a macro for indexing cu_array_t. 
Adds macro CU_ARRAY_AT(cu_array, x, y) to cu.h.
 2016-06-15 12:25:11 +09:00
Arttu Ylä-Outinen 8ac1f1986e Move CU array copy to a separate function. 
Moves code for copying parts of cu_array_t to a new function
kvz_cu_array_copy in cu module.
 2016-06-15 12:25:11 +09:00
Arttu Ylä-Outinen 41e75daed7 Fix overlapping memcpy in kvz_search_cu_smp. 
The destination and source pointers might be equal. Fixed by replacing
the memcpy call with a simple assignment.
 2016-06-15 12:25:11 +09:00
Ari Lemmetti 29af8bcd21 Remove const to match function signature 2016-06-14 18:19:40 +03:00
Eemeli Kallio 5af6ab320c Merge branch 'me_early_terminate' 
Conflicts:
	configure.ac
	src/cfg.c
	src/cli.c
	src/kvazaar.h
	src/search_inter.c
 2016-06-14 15:03:35 +03:00
Eemeli Kallio 43c7778b82 Updated version number. 2016-06-14 10:53:04 +03:00
Arttu Ylä-Outinen 23fdeeaf10 Move mv_cand and mv_dir into a bitfield in cu_info_t. 
Reduces size of cu_info_t.
 2016-06-14 12:21:57 +09:00
Arttu Ylä-Outinen 35aadf6776 Reduce size of type in cu_info_t to two bits. 
Reduces size of cu_info_t.
 2016-06-14 12:21:57 +09:00
Arttu Ylä-Outinen 1cbe844f79 Move inter and intra into an union in cu_info_t. 
Reduces size of cu_info_t.
 2016-06-14 12:21:57 +09:00
Arttu Ylä-Outinen b6d793ef33 Drop field inter.mvd from cu_info_t 
Instead of storing the mv differences in cu_info_t, they are computed
from the mv candidates and the motion vector. Reduces the size of
cu_info_t.
 2016-06-14 12:21:57 +09:00
Arttu Ylä-Outinen 98aa906f30 Drop field coded from cu_info_t 
It can be inferred from the position and size of the CU.
 2016-06-14 12:21:57 +09:00
Arttu Ylä-Outinen ebb10763f1 Drop field inter.mv_ref_coded from cu_info_t. 
Storing inter.mv_ref_coded in cu_info_t is unnecessary since it can be
computed from refmap and inter.mv_ref.
 2016-06-14 12:21:57 +09:00
Arttu Ylä-Outinen 4be5c8f349 Move flags into a bitfield in cu_info_t. 
Reduces the size of cu_info_t.
 2016-06-14 12:21:57 +09:00
Arttu Ylä-Outinen 30e9ee988d Move bitcost field out of cu_info_t.inter. 
The bitcost is only needed for the currently searched CU.

Fixes bitcost of the second PU being ignored when using SMP or AMP.
 2016-06-14 12:21:57 +09:00
Arttu Ylä-Outinen 16d13ed046 Move cost field out of cu_info_t.inter 
The cost is only needed for the currently searched CU.
 2016-06-14 12:20:05 +09:00
Arttu Ylä-Outinen c5c2c182d9 Drop unused field mode from cu_info_t.inter. 2016-06-14 12:18:17 +09:00
Eemeli Kallio e4f1a74512 Added early termination option for motion estimation. 
Conflicts:
	src/search_inter.c
 2016-06-13 16:20:35 +03:00
Wassim Hamidouche 5bc7287c67 add fix for crypro 2016-06-09 10:49:31 +03:00
Wassim Hamidouche 35634b5596 correct MV sign encryption 2016-06-09 10:49:31 +03:00
Wassim Hamidouche 15abdc6e81 correct sign encryption 2016-06-09 10:49:31 +03:00
Wassim Hamidouche 73c3203a26 encry coef transfs 2016-06-09 10:49:31 +03:00
Wassim Hamidouche 7ad5f8bbe5 encry coef transf sign 2016-06-09 10:49:31 +03:00
Wassim Hamidouche 02b0712973 fix g++ compilation 2016-06-09 10:48:44 +03:00
Ari Koivula a2170f0763 Compile the cryptopp wrapper only when used 
This should allow us to avoid an unnecessary dependancy to a C++
compiler.

Conflicts:
	configure.ac
 2016-06-07 17:11:12 +03:00
Ari Koivula 182038c743 Don't allow enabling encryption when it's not compiled in 2016-06-07 16:58:09 +03:00
Ari Koivula 8eb087120e Make VisualStudio ignore the crypto stuff 
Add stubs for the crypto functions so we can refer to them, even if we
never use them.
 2016-06-07 16:58:09 +03:00
Wassim Hamidouche 76cb6dc6c2 add check flags 2016-06-07 10:54:26 +02:00
Ari Koivula 60ea8a359f Add --crypto parameter 2016-06-07 10:31:40 +02:00
Wassim Hamidouche 02308d1ba6 add MVs encryption 2016-06-07 10:28:30 +02:00
Wassim Hamidouche 4637c8a828 compile Kvazaar encoder with ITpp library 2016-06-07 08:33:04 +02:00
Eemeli Kallio 8f182ac6de Added functions select_starting_point and mv_in_merge to search_inter.c 2016-06-06 17:16:04 +03:00
Ari Koivula fe71638a96 Fix problem with ASM compilation 
When compiling C++ files along with C, libtool would complain about
the --tag missing, even though CC should be the default.
 2016-06-06 15:47:56 +03:00
Eemeli Kallio 836a3b1daa Added functions select_starting_point and mv_in_merge. 2016-06-06 12:18:33 +03:00
Ari Koivula 4eaacbe23e Fix bug with lp-gop and ratecontrol 
The first frame was always qp51 due to gop_offset being -1 for the
first frame. This fix makes it so that bits are allocated as if it was
the last (high quality) frame from the previous GOP.
 2016-05-27 15:53:55 +03:00
Ari Koivula 3fbd7ed97f Add GOP layer weights for lowdelay-P 
When using ratecontrol with lowdelay-P, this improves BDRate by 1-25%.
Strongest effect is when using 4 layers and multiple references.

Also allow using 1 or 2 layers with ratecontrol.
 2016-05-27 13:46:26 +03:00
Ari Koivula 67acead4bc Fix referring over IDR boundary when using --gop 
This problem resulted in an illegal bitstream with --gop=lp, because it
uses IDR's. The --gop=8 would not code IDR pictures, even when told to
with -p, which masked this problem.

This fix solves the problem with --gop=lp and also prevents references
across the intra picture in --gop=8. The intra pictures should be set
to IDR in a later fix, or an alternate method of differentiating
between IDR and non-IDR intra should be made.
 2016-05-27 13:20:53 +03:00
Ari Koivula a77dc1610e Refactor encoder_state_remove_refs 
I needed to debug this, so I rewrote it to make sense. There is an
obvious bug with the IDR handling that I left in place to fix in a
separate commit.
 2016-05-27 13:20:45 +03:00
Eemeli Kallio b5c05e58e0 Fixed typo in strategyselector.c 2016-05-24 11:04:29 +03:00
Ari Lemmetti 68c6f0f7b8 Enable deblocking for every preset 
Deblocking adds very little complexity
while giving massive coding performance boost
 2016-05-17 18:50:31 +03:00
Ari Lemmetti 6a07761b46 Add smp and amp options to presets 2016-05-17 14:26:58 +03:00
Ari Lemmetti 3107a93eaf Fix avx2 chroma sampling for amp 2016-05-17 14:09:57 +03:00
Ari Koivula 24d0f9f685 Fix usage message for --hash 2016-05-11 15:03:43 +03:00
Ari Koivula a1c772b696 Merge pull request #136 from MrAsura/cu-split-termination 
Cu split termination

Closes #133.
 2016-05-10 17:22:08 +03:00
Jaakko Laitinen 7010526b1d Removed tabs. 2016-05-10 15:52:44 +03:00
Jaakko Laitinen a77eb5c874 Fixed type conversion error when parsing cu split termination. 2016-05-10 14:34:46 +03:00
Jaakko Laitinen 0d361d5bc7 Moved cu split termination from a pre-processor to a input parameter. 2016-05-10 14:15:41 +03:00
Ari Koivula 1dbe4eb852 Merge branch 'mv-full' 2016-05-10 13:28:07 +03:00
Ari Koivula f6a9d237a3 Merge pull request #134 from miimiz/testink_eemeli 
Strategyselector prints
 2016-05-10 13:27:23 +03:00
Eemeli Kallio 8cfeed852c Added print about SIMD optimizations available and in use to strategyselector. 2016-05-10 12:59:15 +03:00
Ari Koivula f51a68b6fa Add different sizes of search window for full search 2016-04-21 15:11:35 +03:00
Ari Lemmetti efbdc5dade Utilize registers more efficiently for 8x8 and larger blocks 2016-04-21 13:26:38 +03:00
Ari Lemmetti 192cee95b2 Vectorize vertical filtering 2016-04-21 13:26:38 +03:00
Ari Lemmetti 0be35f72b8 Filter 4 pixels simultaneously in x direction 2016-04-21 13:26:38 +03:00
Ari Lemmetti 10484bda9f Make strategies out of fractional pixel sample functions 2016-04-21 13:26:38 +03:00
Ari Koivula 28e7548387 Fix bug in full mv search 
This optimization led to some points not being searched.
 2016-04-21 12:03:57 +03:00
Ari Koivula 2576aeee0b Use merge candidates in full mv search 
Perform a full search window around every mv candidate and the
0-vector.
 2016-04-20 20:47:11 +03:00
Ari Lemmetti 8247faf8e0 Remove 64-bit only instruction to fix 32-bit compilation. 2016-04-19 18:05:11 +03:00
Ari Lemmetti eb55d6b6b9 Fix writing over boundary. 2016-04-19 16:03:43 +03:00
Ari Lemmetti bcabc6fadd Remove pixel blit from strategies. Use memcpy instead. 2016-04-06 18:44:04 +03:00
Ari Lemmetti 2140197ccc Tidy up coeff blit function and use memcpy again. 
Give memcpy constants for fixed sizes to enable copying many bytes simultaneously.
 2016-04-06 18:03:00 +03:00
Ari Koivula 08b4480d94 Re-add time.h include 
Include-what-you-use wants to include sys/time.h instead, or if I
override it to include time.h it will remove the include completely.
 2016-04-02 19:05:16 +03:00
Ari Koivula 61fc3e87ba Run include-what-you-use fix_includes.py fix_includes.py 
The includes should make more sense now and not just happen to compile
due to headers included from other headers.

Used a modified version of IWYU. Modifications were to attribute int8_t
and so on to stdint.h instead of sys/types.h and immintrin.h instead of
more specific headers.

include-what-you-use 0.7 (git:b70df35)
based on clang version 3.9.0 (trunk 264728)
 2016-04-01 17:46:55 +03:00
Ari Koivula 016810d982 Move COMPILE_ macro to global.h 
While these are only used for strategies, it's non-intuitive to have
to include strategyselector.h in every file under strategies before
including anything else.
 2016-04-01 17:46:55 +03:00
Ari Koivula 8908d85d66 Change all relative includes to absolute 2016-04-01 17:46:44 +03:00
Ari Koivula 4876879b82 Add IWYU pragmas 2016-03-31 12:33:34 +03:00
Marko Viitanen 41a5f9bbbe Fix filetime conversion to timespec 2016-03-24 10:08:11 +02:00
Ari Koivula 9139e169fe Fix unnecessary waiting in main thread 
The main thread has to wait for the worker threads to finish. The
pthread_cond_timedwait call used to accomplish this was given
a relative instead of absolute time, which resulted in the call
returning immediately, because the time had already passed.

This removes the now unnecessary sleeps and fixes the time given to
the pthread_cond_timedwait such that it now waits until a job finishes
or 100ms have passed.
 2016-03-23 22:23:04 +02:00
Ari Koivula e23ed231fb Fix race condition with owf and non-square motion partitions 
The OWF wpp limit code assumed square blocks, and as such did not work
correctly when height != width. This changes the relevant code to consider
both height and width.
 2016-03-22 16:46:38 +02:00
Arttu Ylä-Outinen d6a3e02f16 Fix calculating reference CU index in inter search 
Fixes a possible segfault when SMP or AMP blocks are used.
 2016-03-22 12:55:58 +02:00
Ari Lemmetti f4538ab474 Copy pixels more efficiently in lcu recon. 2016-03-18 20:10:03 +02:00
Ari Koivula 5b66578f71 Add kvz_ prefix to md5 functions 
The non kvz_ symbols were being exported in the static lib, which got caught
by Travis tests.
 2016-03-18 13:13:35 +02:00
Ari Koivula 4125218cfa Add --hash=md5 
Add md5 through extras/libmd5 taken from HM with BSD license. It's
implemented as a generic strategy using the same interface as checksum,
so we can write a SIMD version if it seems necessary.
 2016-03-18 05:23:57 +02:00
Ari Koivula 883448b8fb Add --hash parameter 
Allows decoded picture hash to be selected among none and checksum.
 2016-03-18 05:20:15 +02:00
Ari Lemmetti 6d5f8e3aec Define KVZ_COMPILE_ASM for the correct files. 
Enables asm strategies again.
 2016-03-17 16:21:31 +02:00
Ari Lemmetti e502292ba8 Remove old function 2016-03-16 20:18:55 +02:00
Ari Lemmetti c6cc96f5ec Optimize sao band ddistortion 2016-03-16 20:16:00 +02:00
Ari Lemmetti ab577f476f Optimize sao reconstruct color 2016-03-16 20:15:32 +02:00
Ari Lemmetti 48bfddf4ec Optimize calc sao edge dir 2016-03-16 20:14:50 +02:00
Ari Lemmetti ba69992941 Optimize sao edge ddistortion 2016-03-16 20:14:19 +02:00
Ari Lemmetti 941b6b3e27 Optimize calc eo cat 2016-03-16 20:13:30 +02:00
Ari Lemmetti 04fbb48a09 Add strategy for avx2. Copy generic functions there. 2016-03-16 20:13:15 +02:00
Ari Lemmetti 4e30a215d8 Create generic strategy for sao. 2016-03-16 20:11:15 +02:00
Ari Koivula 6f431e510c Comment and tidy threadqueue_worker 
Carefully avoided making any changes to the logic.
 2016-03-14 20:08:04 +02:00
Ari Koivula 1165ae2e1f Increase --mv-constraint=frametimemargin margin 
Increase the margin to be 4 luma pixels to every direction.
 2016-03-14 16:02:54 +02:00
Arttu Ylä-Outinen 0eda28ced6 Fix Visual Studio warnings 
Initialization of a struct with addresses of local variables generated
warning C4221 in encmain.
 2016-03-14 14:12:21 +02:00
Ari Koivula e91ca74733 Refactor kvz_encode_last_significant_xy 2016-03-10 18:47:16 +02:00
Ari Koivula 1fc0e8076c Format kvz_encode_last_significant_xy whitespace 2016-03-10 18:17:45 +02:00
Ari Koivula df9a958ef2 Merge branch 'log2' 2016-03-10 18:16:41 +02:00
Ari Koivula 4112a4364d Remove g_to_bits table 2016-03-10 15:59:51 +02:00
Ari Koivula 9fcfba637f Remove duplicated inline functions 2016-03-10 15:28:31 +02:00
Ari Koivula e27ec2cc53 Add kvz_math.h for common inline math functions 
Calling it just math.h would have prevented including system math.h.
 2016-03-10 15:26:18 +02:00
Ricardo Constantino c515796a21 Only use version prefix in kvazaar binary 
Fixes regression since 54f08f2 causing libkvazaar version checks to not
work (i.e. pkg-config)
 2016-03-09 16:13:59 +00:00
Arttu Ylä-Outinen 54f08f2bdb Use output of git describe as version. 2016-03-09 15:04:29 +02:00
Ari Koivula f8edf28161 Fix const qualifier warning 
Also set the warning to an error in VS.
 2016-03-09 14:16:15 +02:00
Ari Koivula b0c3ece31e Fix race condition when deblocking is on but SAO is off 
Already suspected this yesterday, but didn't want to add the code to
handle it before confirming that it's actually a problem. It is.
 2016-03-09 14:02:46 +02:00
Ari Koivula 1671725c72 Fix non-determinism issue with OWF WPP margin 
The previous reasoning used deblocking and fractional motion estimation
together to arrive at a margin of 4 pixels. This was wrong, and with
either of these off, half pixel chroma interpolation could use pixels
outside the intended region.

Deblocking does not currently affect the margin needed.
 2016-03-08 20:18:38 +02:00
Ari Koivula 674bfa14ce Comment WPP deblocking and SAO 
I was a bit unclear about exactly what happens and when regarding SAO
and deblocking when we do frame-parallel WPP parallelism, so I checked
and commented the bits that were unclear to me.
 2016-03-08 19:39:04 +02:00
Ari Koivula aec152c953 Fix OWF mv restriction limit 
The check was done in regard to the wrong dimension, allowing the
access to unfinished parts of the frame when coding multiple frames
at the same time.
 2016-03-08 17:12:43 +02:00
Ari Koivula fda103aa7c Refactor cfg->tiles_width_count and cfg->tiles_height_count 
Change code everywere so these actually mean "width count" and not
"width count minus one".
 2016-03-07 17:29:15 +02:00
Ari Koivula a350eb3a1e Fix --tiles to have the correct number of tiles. 
The tiles_width_count etc. actually mean "count minus one".
 2016-03-07 17:24:31 +02:00
Ari Koivula 49ea2d7b7f Fix --mv-constraint=frametile 
Option --mv-constraint=frametilemargin was being used instead of
frametile.
 2016-03-07 16:41:00 +02:00
Ari Koivula 95b8dd99f6 Add --tiles parameter 
Add new parameter --tiles that accept only uniform split. I considered
supporting the syntax of --tiles-width-split for this, but writing
--tiles=u2xu2 is just not as intuitive as --tiles=2x2, and there is
hardly ever any reason to use anything but uniform split. The more
cumbersome --tiles-width-split and --tiles-height-split parameters
are still there to allow finer control.
 2016-03-07 16:33:51 +02:00
Ari Koivula fd34dd9bc6 Fix race condition with OWF 
There was an off by one error in the dependance setting code, which
resulted in dependencies not being set resulting in checksum errors.
For example if ref_neg=1 and owf=1.
 2016-03-07 13:38:23 +02:00
Ari Koivula 81b439f4da Optimize starting point selection in tz 
Avoid checking zero motion vectors multiple times. The merge candidate
list often has only one or two candidates, the other being zeroes.
 2016-03-04 16:48:46 +02:00
Ari Koivula 2436702c27 Optimize starting point selection in hexbs 
Avoid checking zero motion vectors multiple times. The merge candidate
list often has only one or two candidates, the other being zeroes.
 2016-03-04 16:48:12 +02:00
Ari Koivula 5327b59b45 Remove KVZ_PERF_SEARCHPX 
It's too invasive and we don't really need it.
 2016-03-04 16:48:12 +02:00
Arttu Ylä-Outinen 348ac4888b Fix calc_mode_bits. 
The CUs left and above the current one would be set to NULL when there
was only one CU between the current one and the left or top edge of the
frame.
 2016-03-04 14:08:35 +02:00
Ari Koivula 86219aa0fc Fix non-determinism with tiles 
Earlier fix that fixed the supply side of the cu_array to take tile
coordinates into account should have been accompanied with this one
that does the same thing to demand side.
 2016-03-03 17:39:20 +02:00
Arttu Ylä-Outinen 626b53ce85 Move sao search from encoderstate to sao. 
Moves sao search from function encoder_state_worker_encode_lcu in
encoderstate.c to function kvz_sao_search_lcu in sao.c. Makes functions
kvz_init_sao_info, kvz_sao_search_chroma and kvz_sao_search_luma static
since they are no longer used outside sao.c.
 2016-03-01 14:56:16 +02:00
Ari Koivula cfa722e448 Reduce parallelism for tiles 
There is still some race-condition with encoding tiles from multiple
frames, so disable this to keep the bitstream deterministic.
 2016-02-29 20:20:21 +02:00
Ari Koivula 3dcc0957f8 Deal with impossible mv constraints 
If 0,0 vector is illegal, it's possible that no legal movement vector,
is found, in which case a large cost is returned instead. The cost
overflowed and there is all sorts of silliness with converting from
double to int, but I'm not going to fix all of it because when we
remove the doubles it will all get fixed.
 2016-02-29 19:18:14 +02:00
Ari Koivula b1adf1576a Add --mv-constraint=frametilemargin 
Add an even stricter motion vector constraint to prevent motion vectors
to fractional pixel positions that would need pixels outside the tile.
 2016-02-29 19:18:14 +02:00
Ari Koivula f808cbf608 Allow increased parallelism for tiles 
When movement vectors are constrained to tiles, only the same tile in
previous frame needs to be depended upon.
 2016-02-29 14:33:06 +02:00
Ari Koivula f4ebff12b0 Combine tile mv constraint with OWF mv constraint 
This also fixes movement vectors in tiles when OWF is on. The OWF mv
constraint assumed WPP, so it didn't work with tiles.
 2016-02-29 14:33:06 +02:00
Ari Koivula 7981609cd0 Add --mv-constraint=frametile 2016-02-29 14:33:06 +02:00
Ari Koivula 9dbbb7fdbc Add --mv-constraint argument 2016-02-29 14:33:06 +02:00
Ari Koivula 1be877faf9 Fix chroma reconstruction with tiles 
An incorrect frame boundary check caused a checksum error, because the
chroma reconstruction of the encoder was wrong. The encoder treated
horizontal tile boundaries as frame boundaries when the vertical
component of the movement vector was a multiple of 8.
 2016-02-29 14:32:51 +02:00